Hiding the Rabbit: Using a genetic algorithm to investigate shape guidance in visual search

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Visual Perception

Background:

• Attention guides visual search using features like shape.
• Current understanding of shape guidance is limited, not fully explaining preattentive processing.
• Novel methods are needed to explore complex shape spaces and guide attention.

Purpose of the Study:

• To develop a novel genetic algorithm to explore shape space and generate hypotheses about shape guidance.
• To investigate how shape similarity influences visual search efficiency.
• To understand the relationship between attended and preattentive shape processing.

Main Methods:

• A genetic algorithm evolved visual search distractors in radial frequency space based on reaction time (RT).
• Observers searched for targets among distractors, with faster RTs promoting distractor survival for easier searches and longer RTs for harder searches.
• The method was tested with both abstract radial frequency shapes and natural silhouettes (e.g., rabbits).

Main Results:

• The genetic algorithm successfully created visual searches that were significantly easier or harder than the initial search within eight generations.
• Easy-to-find distractors evolved dissimilar amplitude x frequency spectra compared to the target.
• Difficult-to-find distractors evolved spectra more similar to the target, and inefficient distractors for natural shapes had unexpected spectral properties.

Conclusions:

• Genetic algorithm-driven evolution of search tasks is a viable method for exploring shape guidance.
• Preattentive shape processing involves similarity in spectral properties, influencing search efficiency.
• Shapes that are distinct when attended can be preattentively similar, complicating visual search.